DARWIN 1.5 : Large Language Models as Materials Science Adapted Learners

Tong Xie, Yuwei Wan, Yuchen Zeng, Clara Grazian, Wenjie Zhang, Dongzhan Zhou, Yixuan Liu, Shaozhou Wang, Chunyu Kit, Wanli Ouyang, Bram Hoex

Keywords: Large Language Models, Materials Science, AI for Material, Foundational Models

TL;DR: DARWIN 1.5 is an open-source LLM for materials science, using multi-task learning and scientific QA fine-tuning to enhance materials property predictions, outperforming GPT-3.5, GPT-4, and traditional ML models.

Confirmation Of Submission Requirements: I submit an abstract. It uses the template provided on the submission page and is no longer than 2 pages.

Submission Number: 194